Transurethral ultrasound diffraction tomography

The potential for cost-effective tomographic imaging using ultrasound continues to be confronted with difficulties arising from the computational complexity of fully three dimensional object reconstruction in the diffraction regime. Development of fast and accurate forward and inverse models for ultrasound propagation in the biomedical frequency range of 1-10 MHz is essential for diffraction tomography to be a practical imaging modality. We have implemented a flexible, object-oriented simulation system in MATLAB for performing rapid two- and three-dimensional modeling of forward scattering using the conjugate gradient FFT method in conjunction with a fast linear adjoint approximation to the Jacobian. Nonlinear conjugate gradient inversion has been implemented and tested in both 2D and 3D, demonstrating the feasibility of the method for diffraction tomography. We have also implemented and tested several regularization schemes including L2-norm and total variation, and have used multigrid iteration in conjunction with anisotropic diffusion filtering to accelerate convergence of the inversion algorithm. Inversions of strongly scattering objects have been successfully performed in 2D and 3D, and results thereof are presented herein.